A Credit Score Is a Prediction, Not a Financial Report Card
A single number on the loan officer's screen — 642 — just added $61 a month to the car payment. Nobody in the room picked it; a model did.
A credit score is a three-digit prediction of how likely you are to repay borrowed money, computed from the data in your credit report. Most common models, including the widely used FICO versions, run from 300 to 850.
What feeds the number
FICO weights payment history most heavily, about 35%, with amounts owed close behind at about 30%. Length of history, new applications, and your mix of account types fill out the rest.
"Amounts owed" leans hard on utilization — balances measured against each card's credit limit. High balances can drag a score down even when every payment lands on time.
Why the price follows the score
Lenders price risk, and the score is their fastest risk estimate. On a $30,000 five-year auto loan, the difference between 7% and 10% is about $43 a month — roughly $2,600 over the loan's life.
That's why the same car, dealer, and paycheck can produce different monthly payments for different buyers. The file walked into the room before you did.
The score isn't the file
The score summarizes the report; the report holds the evidence. Reviewing your reports can surface wrong balances, accounts you never opened, or duplicated negatives — problems the number alone can't explain.
You don't have just one score, either. Different models, bureaus, and data dates produce different numbers, so the score a lender pulls may not match the one in your banking app.
What it can't tell you
A score predicts repayment behavior for the lender's benefit — it isn't a grade on your character or proof that a payment fits your life. Strong credit plus a stretched budget still equals a payment you shouldn't take on.
The habits that build it are unglamorous: pay on time, keep utilization low, keep old accounts open, apply sparingly. Time does the compounding.
The CFPB's credit score explainer covers the mechanics in plain language.
Pull your own reports before any big application — the model reads them whether you have or not.